P300-BCI EEG dataset: sitting vs. supine posture

# P300-BCI EEG dataset: sitting vs. supine posture EEG recordings of a five-target visual P300 brain–computer interface (BCI) task. Ten healthy volunteers performed the task in two postures: sitting (SIT) and supine (SUP). This dataset accompanies the article: > Chailloux Peguero, J.D.; Mendoza-Montoya, O.; Antelis, J.M.; Fabregas, E.; Farias, G.; Velastin, S.A. *Effect of Sitting and Supine Postures on Visual Event-Related Potentials and the Online Performance of a P300-Based Brain–Computer Interface.* Sensors (submitted). Analysis code: <https://github.com/jdchailloux/p300-posture-bci> ## ParticipantsTen healthy volunteers took part: 2 females and 8 males, aged 19–33 years, all with normal or corrected-to-normal vision. They are identified as `sub-01` … `sub-10`, following the participant numbering of the article. `participants.csv` gives, for each participant:- `posture_order`: the order in which the two postures were recorded;- `online_files`: whether online files are available. **sub-02:** the Oz electrode failed during the experiment. Only the offline training files are provided, and the article analysed them without Oz. The study was conducted in accordance with the Declaration of Helsinki. The protocol was approved by the Institutional Research Ethics Committee (CIEI) of the School of Engineering and Sciences, Tecnologico de Monterrey (registration number: CA-EIC-2407-01). All participants gave written informed consent. ## Acquisition- **Amplifier:** g.USBamp with g.SCARABEO active Ag/AgCl electrodes, connected through a g.GAMMAbox driver box (g.tec, Austria).- **Sampling rate:** 256 Hz.- **Hardware filters:** power-line notch filter and 0.5–60 Hz band-pass.- **Electrodes (10–20 system):** Fz, Cz, P3, Pz, P4, PO7, PO8, Oz.- **Reference:** right earlobe. **Ground:** AFz. **Impedance:** below 5 kΩ.- **Units:** microvolts (µV).- **Stimulus display:** 17-inch monitor at 70 cm. In the supine posture, the monitor was mounted on an articulated arm above the participant.- **Software:** BrainTec (gUSBampDataStreamer and P300Analyzer modules). ## ParadigmThe screen showed five items that flashed one at a time in random order. Each flash lasted 75 ms, followed by 75 ms without highlighting (stimulus onset asynchrony of 150 ms). Participants attended to a cued item and silently counted its flashes. - **`run-training` (offline calibration):** 7 blocks. In each block the cued target flashed 40 times, which gives about 280 target and 1120 non-target flashes per file.- **`run-online` (online validation):** 20 selections, made with the subject- and posture-specific classifier. Each selection is cued by a start marker (101–105) and ends with the BCI decision (1001–1005). Each block follows this sequence of phases: | Phase | Start marker | Duration ||---|---|---|| Fixation | 201 | 2 s || Target presentation | 202 | 2 s || Preparation | 203 | 1 s || Stimulation | 101–105 | ≈30 s offline / until selection online || Rest | 200 | 5 s offline / 4 s after an online selection | ## FilesFor each participant, posture (`acq-sit`, `acq-sup`) and run (`run-training`, `run-online`), there are two files. `sub-XX_task-p300_acq-<posture>_run-<run>_eeg.csv` contains the continuous raw EEG (no software processing):- `time_s`: time in seconds from the start of the recording;- 8 EEG channels in µV;- `marker`: event code on the sample where the event occurred, 0 otherwise. `sub-XX_task-p300_acq-<posture>_run-<run>_events.csv` lists the events, one per row:- `onset_s`: time of the event in seconds;- `sample`: sample index, 0-based;- `code`: event code;- `description`: meaning of the code. **Event codes:** | Code | Meaning ||---|---|| 1–5 | flash of item 1–5 || 101–105 | start of stimulation; cued target = code − 100 || 200 | end of stimulation / start of rest || 201 | start of fixation || 202 | start of target presentation || 203 | start of preparation || 1001–1005 | online selection made by the BCI; selected item = code − 1000 | **Target and non-target epochs:** within a stimulation phase, a flash is a *target* if its code equals the cued item (start code − 100). Otherwise it is a *non-target*. **Online metrics:** a selection is correct when the 1001–1005 code matches the cued item. The detection time is the interval between the 101–105 marker and the 1001–1005 marker. ## Notes- The original recordings were stored in the proprietary BrainTec `.ebr` format. They were converted with `export_dataset.m`, which is available with the analysis code.- **Removed during conversion:** the absolute timestamp channel, which contained the date and time of each session.- **Precision:** values are rounded to 4 decimals (0.0001 µV). This is below the single-precision resolution of the original files.- **Not included:** three pilot recordings made at home under different conditions. ## LicenseCreative Commons Attribution 4.0 International (CC BY 4.0).